If W is a fixed, real-valued random variable, then there are simple and easily satisfied conditions under which the function d W , where d W ( X , Y ) = the probability that W “separates” the real-valued random variables X and Y , turns out to be a metric. The observation was suggested by work done in [1].